ReportState

Enumerates the training states reported by the decorator.

mindio_ttp.framework_ttp.ReportState

Parameter	Mandatory/Optional	Description	Value
ReportState	Mandatory	Type of the reported training state. RS_NORMAL: normal state RS_UCE: UCE RS_UCE_CORRUPTED: Multi-bit ECC of the on-chip memory RS_HCCL_FAILED: HCCL recomputation failure RS_UNKNOWN: other errors RS_INIT_FINISH: exception thrown by the newly started ARF node after the training process is initialized in the MindSpore framework RS_PREREPAIR_FINISH: exception thrown by the newly started ARF node RS_STEP_FINISH: exception thrown when the step-level pause in the subhealth hot switchover is complete	RS_NORMAL.value: ttp_c2python_api.ReportState_RS_NORMAL RS_UCE.value: ttp_c2python_api.ReportState_RS_UCE RS_UCE_CORRUPTED: ttp_c2python_api.ReportState_RS_UCE_CORRUPTED RS_HCCL_FAILED.value: ttp_c2python_api.ReportState_RS_HCCL_FAILED RS_UNKNOWN.value: ttp_c2python_api.ReportState_RS_UNKNOWN RS_INIT_FINISH: ttp_c2python_api.ReportState_RS_INIT_FINISH RS_PREREPAIR_FINISH.value: ttp_c2python_api.ReportState_RS_PREREPAIR_FINISH RS_STEP_FINISH: ttp_c2python_api.ReportState_RS_STEP_FINISH

Parameter

Mandatory/Optional

Description

Value

ReportState

Mandatory

Type of the reported training state.

RS_NORMAL: normal state
RS_UCE: UCE
RS_UCE_CORRUPTED: Multi-bit ECC of the on-chip memory
RS_HCCL_FAILED: HCCL recomputation failure
RS_UNKNOWN: other errors
RS_INIT_FINISH: exception thrown by the newly started ARF node after the training process is initialized in the MindSpore framework
RS_PREREPAIR_FINISH: exception thrown by the newly started ARF node
RS_STEP_FINISH: exception thrown when the step-level pause in the subhealth hot switchover is complete

None

Parent topic: API Reference